[Mlir-commits] [mlir] [mlir][mesh, mpi] Lower allreduce (PR #144060)

[Frank Schlimbach]

================
@@ -529,6 +519,124 @@ struct ConvertShardShapeOp : public OpConversionPattern<ShardShapeOp> {
   }
 };
 
+static mpi::MPI_OpClassEnumAttr getMPIReduction(ReductionKindAttr kind) {
+  auto ctx = kind.getContext();
+  switch (kind.getValue()) {
+  case ReductionKind::Sum:
+    return mpi::MPI_OpClassEnumAttr::get(ctx, mpi::MPI_OpClassEnum::MPI_SUM);
----------------
fschlimb wrote:

I'll keep it like this. Extra work introducing an indirection just to make the lines a few characters doesn't look right to me.

https://github.com/llvm/llvm-project/pull/144060